How to evaluate your portfolio (beyond the return)

Two portfolios, one lesson

Take two portfolios almost anyone might hold. Portfolio A is “just own the winners” — an equal-weight basket of the Magnificent Seven (Apple, Microsoft, Nvidia, Alphabet, Amazon, Meta, Tesla). Portfolio B is a boring diversified blend — a chunk of the S&P 500, a few quality individual names, plus bonds, gold, energy and international stocks. We scored both over the same three years against the S&P 500 (SPY). Here is what came back.

Look at the return row alone and Portfolio A is the obvious genius: 37% a year versus 22%. Now look at the Sharpe row. Portfolio A’s Sharpe is 1.36 — essentially identical to just buying the index (1.35). All that extra return came with proportionally more risk: 26% volatility and a stomach-churning -29% drawdown. On a risk-adjusted basis, the “winner” didn’t beat a plain index fund at all. The quiet diversified portfolio, meanwhile, posted the best Sharpe of the three (1.68) and the shallowest drawdown, while roughly matching the index’s return.

The scatter says it in one picture. Up is more return; right is more risk. Portfolio A flies high but far to the right — it bought its returns with volatility. Portfolio B sits up and to the left of the index: a touch more return for less risk, which is the only free lunch in investing. This is why “what did you return?” is the wrong first question. The right one is “what did you return per unit of risk?”

The six numbers that matter

1. Risk-adjusted return (Sharpe and Sortino)

The Sharpe ratio is return divided by volatility — reward per unit of risk. The Sortino ratio is its smarter cousin, dividing by downside volatility only, since upside swings aren’t the kind of risk that hurts. Rough guide: under 1 is unremarkable, ~1 is solid, over 2 is excellent and hard to sustain. But the absolute number matters less than the comparison: beat the index’s Sharpe, or you were just taking more risk for the same money.

2. Maximum drawdown (and Calmar)

Max drawdown is the worst peak-to-trough fall you would have lived through. It is the number that decides whether you actually hold a strategy or panic-sell at the bottom. A backtest that returns 30% a year means nothing if it asked you to sit through a 50% loss; most people can’t. The Calmar ratio (CAGR divided by max drawdown) rewards return that didn’t come with a near-death experience. Portfolio A’s -29% drawdown is almost double Portfolio B’s -15%.

3. Beta and correlation

Beta says how much your portfolio amplifies the market: Portfolio A’s 1.48 means it moved ~1.5x the index, both ways. Correlation says how tightly it tracks the market at all. High beta plus high correlation is a quiet confession: your portfolio is mostly the market with the volume turned up. That’s fine if you want it — but don’t mistake it for stock-picking skill. Real skill shows up as alpha: return left over after subtracting what the market gave you. Both example portfolios earned similar alpha (~+4.8%); B just did it far more efficiently.

4. Concentration — and its look-through trap

The Herfindahl index (HHI) and top-position weights measure how much your eggs share a basket. But here’s a subtlety the raw number hides: both example portfolios score a similar HHI (~0.14), yet Portfolio B is far less concentrated in reality — because its biggest “position” is an S&P 500 fund holding 500 companies. Concentration measured on tickers is blind to look-through. A single broad ETF can be more diversified than ten hand-picked stocks. Read the number, then sanity-check what’s actually inside.

5. Sector exposure

Your portfolio can be diversified across names and still be a single bet. Portfolio A holds seven different companies — and effectively one theme: large-cap tech and AI. When the exposure report shows every holding clustered in a handful of related sectors, your “diversification” is cosmetic. One macro shock hits all of it at once.

6. Risk contribution — where the risk actually lives

This is the metric almost no one looks at, and it’s the most revealing. A position’s share of your money is not its share of your risk. In the equal-weight Mag7 portfolio, every name is ~14.3% of the capital — but they are nowhere near 14.3% of the risk.

Tesla is one-seventh of the money but nearly a quarter of the risk; Tesla and Nvidia together are under a third of the weight yet contribute over 40% of the volatility. If you only watch position sizes, you have no idea that two names are quietly driving the whole portfolio’s fate. Risk contribution — computed from the covariance of returns — is what surfaces it.

How to run this on your own portfolio

You don’t compute covariance matrices by hand. With an agent connected to a markets database over MCP, you hand it your holdings and ask for the block. The ad-hoc tool runs the full analysis without saving anything — ideal for testing a candidate basket before you commit to it:

“Analyze this portfolio over three years vs SPY: 30% SPY, 8% each AAPL, MSFT, JNJ, JPM, XOM, GLD, TLT, VEA, 6% NVDA. Give me CAGR, volatility, Sharpe, Sortino, max drawdown, beta, correlation, alpha, sector exposure, concentration and per-position risk contribution.”

That single request returns every number in the table above. From there you can ask the optimizer to suggest weights that maximize Sharpe or minimize volatility, or to cut the risk a single position is contributing. The point isn’t to chase a perfect score — it’s to see your portfolio the way a risk desk would, instead of through the flattering lens of its best line: the return.